1. Field of the Invention
The present invention relates to a moving pictures encoding technology, and particularly to a low-delay encoding method capable of reducing a time delay generated in encoding and decoding operations into a boundary needed in an actual application.
2. Description of Related Art
Recently, systems such as a high-definition television (HD-TV), a high definition video cassette recorder (HD-VCR), a digital camcorder, a multimedia videophone, a video conference unit, chiefly use a method for digitally encoding and decoding moving pictures and associated audio. The MPEG (Moving Pictures Experts Group) for international standardization of the encoded representation of moving pictures and associated audio, classifies patterns of the pictures into an intra-coded (I) picture, a predictive-coded (P) picture and a bidirectionally predictive-coded (B) picture according to three different encoding methods, and proposes a GOP (Group of Pictures) layer for encoding video data having a picture sequence of "IBBPBBPBBI.multidot..multidot..multidot.." The schemes for encoding and decoding the video data using the GOP structure proposed by the MPEG is described in "ISO/IEC 13818-2." A special playback function such as random access function can be accomplished with respect to the video data encoded using the GOP structure in the decoding end. However, since the decoding end cannot perform a normal decoding operation with respect to the GOP until all the video data of one GOP has been received, a decoding delay phenomenon occurs. Such a decoding delay phenomenon occurs when the decoding end starts to decode a video bitstream which is newly applied due to a channel change, power-on or reset operation.
To reduce a decoding delay time due to the above-described GOP structure, a low-delay encoding technology for encoding video data having a picture sequence of "IPPP.multidot..multidot..multidot." is being researched and developed. Since the structure of "IPPP.multidot..multidot..multidot." does not adopt a concept of the GOP, the random access function is not possible. However, there are merits that a decoding delay time is shorter than that of the GOP structure and complexity of hardware is reduced, because of absence of the B-picture. The low-delay encoding method intra-codes only a portion of the data in each frame, unlike the GOP structure, which adopts a forced intra-frame encoding method with respect to a total of one picture. There are an intra-slice method and an intra-column method as encoding methods used in a low-delay encoding technology. The intra-slice method intra-codes a predetermined number of slices or one slice per picture from the uppermost end to the lowermost end with respect to all P-pictures. FIGS. 1A and 1B show movement of the slice intra-coded by the intra-slice method. On the other hand, the intra-column method intra-codes a column composed of a unit of a predetermined number of bits per picture from the leftmost end to the rightmost end with respect to all P-pictures. FIGS. 2A and 2B show movement of the column intra-coded by the intra-column method.
However, it is unknown that the intra-slice method is not proper in case of a large amount of vertical movement, and intra-column method is not proper in case of a large amount of horizontal movement. Thus, using only one method cannot completely reconstruct the picture. Problems will be described below, which occur when the video data encoded by the intra-column is applied to a receiver. If a channel change or power-on operation is performed, a decoder in the receiver stores video data in the intra-coded image region and video data in an inter-coded image region as a reference frame picture, and performs motion compensation to restore the picture of the current frame. In this case, however, the reference frame data used for the motion compensation to restore the current frame picture includes the predictive-coded P-picture data. Accordingly, in case of a large amount of the horizontal movement, artifacts are not completely removed from the restored picture even after the intra-coded columns corresponding to one entire screen have been inputted. Such a problem occurs in the same manner when the data encoded by the intra-slice method is restored and in case of a large amount of motion in the vertical direction, as well.